Don't use VertxRecorder to get Vertx instance

[geoand]

No description provided.

[vsevel]

can't validate the change, but we are going to fetch your branch and try on a project internally. thanks!

[geoand]

👌🏼

[kdubb]

@geoand Please look at #109 as I referenced in the other thread. This is almost exactly how I originally wrote it and it was switched to using VertxRecorder in #109 (you opened the PR 😏)

[vmargerin]

@vsevel Changes validated on my side

[geoand]

Oh right... I have no memory of that whatsoever

[kdubb]

Maybe something has changed in 3.4 that allows that to work again?

Here is your original problem statement for #78...

Essentially the problem is that when creating the synthetic bean for the PgPool, Quarkus loads the CredentialsProvider (via CDI) which then attempts (in this case) to load Vertx from CDI. This last load will fail because Vertx is also a synthetic bean.

[kdubb]

I thought this use case had test coverage via quarkus. If it doesn't, it should be added to ensure this doesn't regress as we attempt to change this.

[geoand]

It probably works again by luck.

However I do know what is needed to fix it - I might have time to address the deep problem I described in #78 in Quarkus.

[geoand]

It turns out that the change is breaking

[geoand]

quarkusio/quarkus#35949 should fix things once and for all. With that change, one would be able to obtain Vert.x from CDI without any timing issues creeping in.

[vsevel]

regardless of wether or not quarkusio/quarkus#35949 (comment) gets backported to 3.4, I think we need to go ahead with that fix. ok for you @kdubb ?

[geoand]

If worse comes to worse, you can always temporarily use VertxEventBusConsumerRecorder instead of VertxRecorder

[vsevel]

you can always temporarily use VertxEventBusConsumerRecorder instead of VertxRecorder

@geoand so that would be an interim fix for 3.4, and we get your fix for 3.5?
it comes down to whether or not quarkusio/quarkus#35949 is going to be backported @gsmet @cescoffier ?

[geoand]

@vsevel correct on both counts

[kdubb]

@vsevel Go ahead with which fix? This PR? I have trouble accepting a PR that fixes the 3.4 build but we know doesn't work with 3.4. This is near and dear to myself as we use this feature we would be breaking.

IIUC It sounds like we can change this PR to use VertxEventBusConsumerRecorder and then if/when #35949 is merged/backported we then just revert this PR and & #109.

[gsmet]

Please use VertxEventBusConsumerRecorder for now so that we can get a release compatible with 3.4 before I release the Platform tomorrow around noon.

[vsevel]

@gsmet I will work on this tomorrow morning

[vsevel]

this PR will continue in the context of quarkus 3.5

[vsevel]

@geoand is this the plan to upgrade to CR1 today and finalize this PR?
do you need some help?

[geoand]

Sure yeah, I can update the PR once CR1 is out